Metabolomic signature of exposure and responseto citalopram/escitalopram in depressedoutpatientsSudeepa Bhattacharyya, University of ArkansasAhmed T. Ahmed, Mayo ClinicMatthias Arnold, Duke UniversityDuan Liu, Mayo ClinicChunqiao Luo, University of ArkansasHongjie Zhu, SanofiSiamak Mahmoudiandehkordi, Duke UniversityDrew Neavin, Mayo ClinicGregory Louie, Duke UniversityBoadie Dunlop, Emory University

Only first 10 authors above; see publication for full author list.

Journal Title: Translational PsychiatryVolume: Volume 9, Number 1Publisher: Springer Nature [academic journals on nature.com]: Fully openaccess journals | 2019-07-04, Pages 173-173Type of Work: Article | Final Publisher PDFPublisher DOI: 10.1038/s41398-019-0507-5Permanent URL: https://pid.emory.edu/ark:/25593/tvf11

Final published version: http://dx.doi.org/10.1038/s41398-019-0507-5

Copyright information:© 2019, The Author(s).This is an Open Access work distributed under the terms of the CreativeCommons Attribution 4.0 International License(https://creativecommons.org/licenses/by/4.0/).

Accessed February 20, 2022 6:56 PM EST

Bhattacharyya et al. Translational Psychiatry (2019) 9:173

https://doi.org/10.1038/s41398-019-0507-5 Translational Psychiatry

ART ICLE Open Ac ce s s

Metabolomic signature of exposure andresponse to citalopram/escitalopram indepressed outpatientsSudeepa Bhattacharyya 1, Ahmed T. Ahmed2, Matthias Arnold 3,4, Duan Liu 5, Chunqiao Luo1, Hongjie Zhu6,Siamak Mahmoudiandehkordi3, Drew Neavin5, Gregory Louie 3, Boadie W. Dunlop7, Mark A. Frye2, Liewei Wang5,Richard M. Weinshilboum5, Ranga R. Krishnan8, A. John Rush3,9,10 and Rima Kaddurah-Daouk 3,11,12

AbstractMetabolomics provides valuable tools for the study of drug effects, unraveling the mechanism of action and variationin response due to treatment. In this study we used electrochemistry-based targeted metabolomics to gain insightsinto the mechanisms of action of escitalopram/citalopram focusing on a set of 31 metabolites from neurotransmitter-related pathways. Overall, 290 unipolar patients with major depressive disorder were profiled at baseline, after 4 and8 weeks of drug treatment. The 17-item Hamilton Depression Rating Scale (HRSD17) scores gauged depressivesymptom severity. More significant metabolic changes were found after 8 weeks than 4 weeks post baseline. Withinthe tryptophan pathway, we noted significant reductions in serotonin (5HT) and increases in indoles that are known tobe influenced by human gut microbial cometabolism. 5HT, 5-hydroxyindoleacetate (5HIAA), and the ratio of 5HIAA/5HT showed significant correlations to temporal changes in HRSD17 scores. In the tyrosine pathway, changes wereobserved in the end products of the catecholamines, 3-methoxy-4-hydroxyphenylethyleneglycol and vinylmandelicacid. Furthermore, two phenolic acids, 4-hydroxyphenylacetic acid and 4-hydroxybenzoic acid, produced throughnoncanconical pathways, were increased with drug exposure. In the purine pathway, significant reductions inhypoxanthine and xanthine levels were observed. Examination of metabolite interactions through differential partialcorrelation networks revealed changes in guanosine–homogentisic acid and methionine–tyrosine interactionsassociated with HRSD17. Genetic association studies using the ratios of these interacting pairs of metaboliteshighlighted two genetic loci harboring genes previously linked to depression, neurotransmission, orneurodegeneration. Overall, exposure to escitalopram/citalopram results in shifts in metabolism through noncanonicalpathways, which suggest possible roles for the gut microbiome, oxidative stress, and inflammation-relatedmechanisms.

IntroductionMajor depressive disorder (MDD) is a common, often

disabling condition affecting over 300 million individuals

worldwide1. Selective serotonin reuptake inhibitors(SSRIs) are common first-line treatments for MDD2,3.They are believed to increase the extracellular availabilityof the neurotransmitter serotonin by limiting its reab-sorption into the presynaptic cell, so that serotonin levelsare increased in the synaptic cleft and available forbinding to postsynaptic receptors. Responses to anti-depressant medications are modest. Only about half thepatients respond to the first medication; only one in three

© The Author(s) 2019OpenAccessThis article is licensedunder aCreativeCommonsAttribution 4.0 International License,whichpermits use, sharing, adaptation, distribution and reproductionin any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if

changesweremade. The images or other third partymaterial in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to thematerial. Ifmaterial is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtainpermission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Correspondence: Rima Kaddurah-Daouk (rima.kaddurahdaouk@dm.duke.edu)1Department of Biomedical Informatics, University of Arkansas for MedicalSciences, Little Rock, AR, USA2Department of Psychiatry and Psychology, Mayo Clinic, Rochester, MN, USAFull list of author information is available at the end of the article.These authors contributed equally: Ahmed T. Ahmed, Matthias Arnold

1234

5678

90():,;

1234

5678

90():,;

1234567890():,;

1234

5678

90():,;

achieves symptom remission, which is the virtual absenceof symptoms and the aim of treatment4. Some patients dowell on a single medication, while others require medi-cation combinations or alternative interventions. Clinicalsymptoms are insufficient to guide appropriate treatmentselection5 and, presently, treatments are therefore selectedempirically relying on a “trial and error” approach6,7.Metabolomics, a promising new approach to under-

standing depression and other neuropsychiatric dis-orders8–11, could help inform treatment selection12,13.Metabolomic profiles provide informative readouts onpathways and biological networks implicated in variousdiseases or their treatments. Metabolomic signatureshave been identified for several psychiatric disorders,such as MDD14, bipolar disorder15,16, and schizo-phrenia17–19. Most studies of mood disorders haveimplicated tryptophan (TRP), tyrosine, and purinemetabolism, since historically, neurotransmission andserotonergic signaling were key focus areas of investi-gation14. The TRP pathway along with its three branchesof metabolism to serotonin/melatonin/5-hydro-xyindoleacetate, kynurenine (KYN), and indole deriva-tives, seems to be affected in the depressed state20–28.The purine pathway, whose regulation seems to beconnected to TRP metabolism, has also been implicatedin depression and other psychiatric disorders29. Amongpatients in remission from a major depressive episode, ametabolomic signature that included methionine, glu-tathione along with metabolites in the purine and TRPpathways, has been identified30.Pharmacometabolomics has also revealed that patients’

metabolomic profiles (metabotypes), both prior to andearly during treatment, can inform treatment out-comes10,31. This approach has been applied to anti-hypertensive32 and antiplatelet33 therapies. We have usedthis approach to predict treatment outcomes and toidentify specific metabolomic pathways that were changedin response to sertraline34,35 and to ketamine36, a pro-mising agent for treatment-resistant depression. We havealso employed a “pharmacometabolomics-informedpharmacogenomics” research strategy11 to investigate therole of genetics in response to citalopram or escitalo-pram37,38, thereby advancing the goal of precision medi-cine for depression31. However, the acute and longer-termeffects of treatment with citalopram or escitalopram onpathways, critical to the pathobiology or pharmacotherapyof depression, and the relationship to clinical outcomeshave not been reported.This report used metabolomic analyses with selected

metabolites in the tryptophan, tyrosine, purine, toco-pherol, and the related pathways in a sample of non-psychotic depressed outpatients who were treated for8 weeks with citalopram or escitalopram to address thefollowing questions:

– The metabolomic signature of exposure toescitalopram/citalopram: which metabolite changesoccurred from baseline to week 4, and from baselineto week 8 of treatment?

– The metabolomic signature of response: whichmetabolomic changes were related to changes indepressive symptoms (HRSD-17), longitudinally, inthe overall population and also in responders versusnonresponders?

– The interrelationships between metabolites: what arethe relationships among metabolites, both withinand between pathways, before and after treatmentwith the drug?

MethodsStudy design and participantsWe used samples from the Mayo Clinic NIH-

Pharmacogenomics Research Network-AntidepressantPharmacogenomics Medication Study (PGRN-AMPS)which recruited a total of 803 MDD patients39. Patientselection, symptomatic evaluation, and blood samplecollection for the PGRN-AMPS clinical trial have beendescribed elsewhere24,38–40. Briefly, MDD patients wererequired to have a baseline HRSD17 score ≥ 14, and allpatients who completed 8 weeks of treatment (n= 290)were treated with one of the two SSRIs, citalopram orescitalopram. Depressive symptoms were assessed withHRSD17 at baseline, week 4, and week 8 of SSRI treat-ment. Blood samples were collected at these same timepoints.The HRSD17 was used to ascribe “response”—defined as

at least 50% reduction in the total score from baseline toexit; “remission” —an exit HRSD17 score of 7 or less; and“complete-non-response”—less than 30% reduction in theHRSD17 total score from baseline to exit39. Genome-wideassociation studies for plasma concentrations of the SSRIsand metabolite levels40 and for response41 in this trialhave been published previously. The trial was designed asa parallel to the large National Institute of Mental Health—funded “the Sequenced Treatment Alternatives toRelieve Depression” (STAR*D) clinical trial42 for thepurpose of replication of the identified genetic markers.

Metabolomic profilingA targeted, liquid chromatography–electrochemical

coulometric array (LCECA) metabolomics platform43 wasused to assay metabolites in plasma samples from thethree time points, baseline, 4 weeks, and 8 weeks. Thisplatform was used to identify and quantify 31neurotransmitter-related metabolites (against standards)primarily from the TRP, tyrosine, and tocopherol path-ways, including serotonin. A list of the metabolites thatwere quantitatively measured using this platform is pre-sented in Table 1.

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 2 of 14

Analysis methodThe long-gradient LCECA method used for this analysis

can resolve compounds at picogram levels through elec-trochemical detection (resulting from oxidation orreduction reactions) including multiple markers of oxi-dative stress and protection. This method utilizes a 120-min gradient from (0%) organic modifier with an ion-pairing agent (i.e., pentane sulfonic acid) to a highlyorganic mobile phase with methanol (80%)/isopropanol(10%)/acetonitrile (10%). An array of 16 serial coulometricelectrochemical detectors is set at incremental potentialsfrom 0 to 900mV, responding to oxidizable compoundssuch as tocopherol in lower potential sensors and higheroxidation potential compounds such as hypoxanthine inthe higher potential channels.

Analysis sequence and data outputAt the time of preparation, a pool was created from

small aliquots of each sample in the study, which was thentreated identically to a sample. All of these assays wereexecuted in sequences that included mixed standard, five

samples, pool, five samples, mixed standard, and so onand so forth. In this study, all sample run orders wererandomized. The sequences decreased possible analyticalartifacts during further data processing. Data were timenormalized to a pool at the midpoint of the study, aligningmajor peaks to 0.5 s and minor peaks to 0.5–2 s. Detailson the LCECA methods are described in previouslypublished work35,41,44–50.

Data analysisAll data preprocessing and analysis were performed

with R (version 3.4.2) and Bioconductor (version 3.3)statistical packages.

PreprocessingThis study’s data extraction protocol followed the

STORBE guidelines46. All metabolite data were firstchecked for missing values (none were detected at >20%missing abundances) and were subjected to imputation bythe k-nearest neighbor algorithm51. Data were then log2

Table 1 List of metabolites and pathways analyzed in the study

Metabolite by pathways Abbreviation Metabolite by pathways Abbreviation

Tryptophan Phenylalanine/tyrosine

3-Hydroxykynurenine 3OHKY 4-hydroxybenzoic acid 4HBAC

5-Hydroxyindoleacetic acid 5HIAA

5-Hydroxytryptophan 5HTP Purine

Indole-3-acetic acid I3AA Guanine G

Kynurenine KYN Guanosine GR

Serotonin 5HT Hypoxanthine HX

Tryptophan TRP Uric acid URIC

Xanthine XAN

Tyrosine Paraxanthine PXAN

4-Hydroxyphenylacetic acid 4HPAC Xanthosine XANTH

4-Hydroxyphenyllacetic acid 4HPLA

Homogentisic acid HGA 1 Carbon+ GSH

Homovanillic acid HVA Methionine MET

Methoxy-hydroxyphenyl glycol MHPG Cysteine CYS

Tyrosine TYR

Vanillylmandelic acid VMA Other

Salicylate SA

Tocopherol Alpha-Methyltryptophan AMTRP

Tocopherol-alpha ATOCO Indole-3-propionic acid I3PA

Tocopherol-delta DTOCO Theophylline Theophylline

Tocopherol-gamma GTOCO

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 3 of 14

transformed and scaled to unit variance prior to statisticalanalyses.

Univariate analysisTo define the effect of drug exposure over 4 weeks and

8 weeks of treatment, linear mixed effects models (usingthe R package nlme52) were fitted on each metaboliteadjusting for age, gender, and HRSD17 scores at baselinewith subjects as random variable. Analyses were con-ducted separately for 4 and 8 weeks. Linear mixed effectsmodels were also used to determine associations betweenthe changes in metabolites and changes in HRSD17 overtime, with age and gender as covariates, and using sub-jects as random variable. All p-values were used to cal-culate the false discovery rates by Benjamini–Hochbergmethod53, and a cutoff point of 10% was used. A two-stepregression strategy was used to find metabolites withsignificant temporal changes and significant differencesbetween responders and nonresponders using themaSigPro library in R54. First, a least-squared techniquewas employed to identify differential metabolites in aglobal regression model, using dummy variables forexperimental groups. Second, stepwise regression wasapplied to select variables that differed between theexperimental groups and find significantly differentmetabolite profiles between the groups.

Partial correlation networks with cluster subgraph analysisThe relationship between metabolites in a complex

disease setting can be represented in terms of partialcorrelation networks, where each node represents ametabolite and each edge between two metabolitesrepresents that two variables are not independent afterconditioning on all variables in the dataset. These edgeshave a weight, edge weights, which are the partial corre-lation coefficients. Here, we estimated the partial corre-lation matrix for all of the metabolites using the leastabsolute shrinkage and selection parameter (LASSO) toobtain the sparse inverse covariance matrix to avoidoverfitting and spurious correlations. Thus, it can bereasonably expected that the regularized partial correla-tion networks will provide accurate estimates of theunderlying relationships between the metabolites inmetabolic pathways and reactions. The LASSO regular-ization parameter was set via EBIC or Extended BayesianInformation criterion50. Finally, the walktrap algorithm,which is based on random walks to capture clusterstructures in a network, is used to identify clusters ofstrongly interacting metabolites45. The final network withcluster subgraphs is formed by the median pairwise partialcorrelations over 1000 bootstrap estimations and plottedusing the Fruchterman–Reingold layout. We furtherincluded the HRSD17 scores in our partial correlationnetwork models to perform differential network analysis.

The overall statistical impact of HRSD17 scores on themetabolite interactions was calculated based on measur-ing structure invariance between two networks, highHRSD17 and low HRSD17 networks, constructed using amedian split of the variable. Permutation tests were usedto determine the significance of structure and edgeinvariances between the two networks55. Themetabolite–metabolite partial correlations that were ofdifferential strength between networks of high and lowHRSD17 networks were further validated for significantinteraction effects through linear regression analysis.

Candidate metabolic trait GWAS with HGA/GR and MET/TYRratiosFor 288 of the 290 subjects in this study we had geno-

type data for the Illumina human 610-Quad BeadChips(Illumina, San Diego, CA, USA) available, as describedpreviously38,41. Genotype QC using PLINK and imputa-tion followed standard protocols. Briefly, raw genotypedata were filtered for variants with call rate <5%, minorallele frequency (MAF) <5%, and Hardy–Weinberg equi-librium HWE p < 1 × 10−549. The data was then subjectedto prephasing using SHAPEIT2 (ver. 2.12)48, followed byimputation with IMPUTE2 (ver. 2.3.2)47 using 1000 gen-omes phase 3 version 556 haplotypes as a reference. Post-imputation QC included filtering variants for IMPUTEinfo score < 0.5, call rate and MAF < 5%, and HWE p < 1 ×10−5, resulting in a final set of 5.55 mio SNPs with 99.14%genotyping rate. To remove any potential for spuriousassociations due to population stratification, we used a setof about 100,000 SNPs pruned for the LD structure andretrieved the first five principal component eigenvectors(PCs). Metabolite data for the HGA/GR and Met/TYRratios were log transformed, centered to zero mean, andscaled to unit variance. In addition, for candidate GWAS,we excluded values that were more than 4 standarddeviations from the mean. We then performed GWAS forHGA/GR and MET/TYR at each time point whileadjusting for age, sex, and PCs 1–5. We reran the GWASadditionally adjusting for HDRS17 scores at each timepoint to eliminate the effects linked to depression severity.

ResultsPatient characteristicsPlasma metabolite data were available from 290 MDD

patients. The average age of the patient cohort was 39.8(±13.1) years. Females comprised of 66% of the studycohort, while males were at 34%. The response rate to thedrug, based on HRSD17 scores, was 69.3% after 8 weeks,compared with 30.7% who were classified as non-responders for this study. The depressive status of thepatients, as determined by the HRSD17 scores, decreasedover time with the drug treatment, from an average of21.9 (±4.9) at baseline to 11.6 (±6.4) at week 4 and 8.6

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 4 of 14

(±5.5) at week 8. Demographic and clinical characteristicsare detailed in Supplemental Table 1.

Metabolite changes at weeks 4 and 8 compared withbaseline, in response to the drugSeveral metabolites in the purine, tryptophan, and tyr-

osine pathways changed, following 4 weeks of drug ther-apy. However, perturbations in the metabolite levels werein general, greater and more significant after 8 weeks oftreatment (Supplemental Table 2). Figure 1 illustrateschanges within key pathways evaluated after 8 weeks oftreatment.

Tryptophan pathwayDramatic changes were observed in serotonin (5HT)

and the ratio 5HIAA/5HT, both at week 4 and week 8. Atboth time points, 5HT showed substantial decreases andthe 5HIAA/5HT ratio was significantly elevated. WhileTRP itself did not show a notable change, its indole-containing metabolite I3AA was significantly elevated, aswas the ratio of I3AA/TRP, possibly indicating a shift

away from the serotonergic pathway of TRP metabolism.Interestingly, another indole-containing compound that isknown to be produced only by gut microbiota in humans,I3PA, was also increased at 8 weeks (unadjusted p-value <0.02). No statistically significant alterations were observedin the KYN branch of TRP metabolism.

Tyrosine pathwayA similar trend of a shift to noncanonical branches of

tyrosine metabolism was also observed in this pathway.MHPG, the major metabolite of the neurotransmitternorepinephrine and the ratio MHPG/TYR showed sig-nificant reductions in their blood levels at both 4 and8 weeks while VMA, a norepinephrine end metabolite,showed significant elevations at 8 weeks compared withbaseline. A phenolic acid, 4HPAC, and its ratio to TYR(4HPAC/TYR) were significantly increased at both 4 and8 weeks. Another phenolic derivative from the phenyla-lanine/tyrosine pathway, 4-hydroxybenzoic acid(4HBAC), was also significantly elevated at 8 weeks.

Fig. 1 Metabolic signature of drug exposure. a Shows the heatmap of metabolite changes at baseline, week 4, and week 8, normalized to baselinelevels. b–d Show changes within the purine, tryptophan, and tyrosine pathways

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 5 of 14

Purine pathwayThe purine metabolites HX and XAN and the ratio

XAN/XANTH were decreased significantly, while theratios PXAN/XAN and URIC/XAN were elevated at8 weeks compared with baseline, indicating a similardecline in the canonical pathway of purine metabolism, asobserved in the tryptophan and tyrosine pathways.Other metabolites that showed significant changes,

albeit at unadjusted p values <0.05, were the purinemetabolites, G, PXAN, and XANTH; the TRP metabolite,5HTP; the tyrosine metabolite, HGA; and other metabo-lites, such as salicylic acid (SA).

Metabolomic changes associated with changes indepressive symptoms (HRSD17)Using linear mixed models, we examined the associa-

tion between temporal changes in metabolite levels(across three time points, baseline, 4 weeks, and 8 weeks)and the temporal changes in patients’ HRSD17 scores over

that period of time (see Fig. 2a–c and Supplemental Table3). In the overall population, metabolites from the TRPpathway were associated with changes in HRSD17 scores.5HT, 5HIAA and the serotonin turnover marker 5HIAA/5HT showed significant positive and negative associa-tions, respectively, with decreases over time in HRSD17

scores (FDR-adjusted p values <0.01).We further subcategorized the population based on

their HRSD17 scores after 8 weeks of treatment. If theyhad at least a 50% reduction in their HRSD17 scores, frombaseline to exit, they were categorized as responders,otherwise they were nonresponders. We examined whe-ther the temporal associations between metabolite chan-ges and HRSD17 scores significantly differed betweenresponders and nonresponders. The mean (±sd) HRSD17

scores in the responders and nonresponders were 21.86(±5.17) and 22.03 (±4.28), respectively, at baseline, 10.10(±5.77) and 15.03 (±6.58), respectively, at week 4, and 5.79(±3.27) and 14.90 (±4.15), respectively, at week 8. 5HT

Fig. 2 Metabolite changes associated with HRSD17 scores. a 5HT (Serotonin), b 5HIAA/5HT ratio, and c 5HIAA (5-hydroxyindoleacetic acid).Temporal changes in (d) HRSD17 scores, (e) 5HT, and (f) MHPG differed significantly between responders and nonresponders. The error bars representstandard error of the mean

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 6 of 14

temporal profiles significantly differed between the twogroups, with the levels being consistently higher in theresponders at baseline, week 4, and week 8, while thedecline in HRSD17 scores was significantly lower at both 4and 8 weeks compared with baseline (Fig. 2d, e). Levels ofMHPG at baseline were significantly higher and the dropin MHPG levels over time was significantly greater inresponders compared with nonresponders (Fig. 2f).

Relationships amongst metabolites at baseline and after8 weeks of treatmentBiological systems are now increasingly viewed as

complex networks of interlinked entities, topologicalanalyses of which can reveal the underlying landscape ofbiological functionalities. Gaussian graphical modelinghas been used to reconstruct pathway reactions in meta-bolomics data57. Combining a partial correlation networkand genetic variation through GWAS has been shown toprovide an in-depth overview of the underlying mechan-istic pathways58. Here, using regularized partial correla-tion network analysis at baseline and also after week 8 ofdrug exposure (Fig. 3), we assessed themetabolite–metabolite interactions between tryptophan,tyrosine, purine, and tocopherol pathways.Regularized partial correlation networks of the meta-

bolites at baseline (Fig. 3a, b) and also at week 8 (Fig. 3c,d) showed significant correlations between several meta-bolites, both within and between pathways forming clus-ters of interacting molecules. A list of statisticallysignificant partial correlations between metabolites atbaseline and week 8 are presented in Supplemental Table4 A, B. Important observations through cluster subgraphanalysis showed that MET, TYR, and TRP formed a tightcluster both at baseline and week 8. However, the strengthof interactions between MET and TYR was significantlyreduced at week 8, compared with baseline (~50%reduction, permutation p value < 0.10). GR connection tothis cluster was significant at week 8 through interactionswith all three metabolites. HVA formed a significantcorrelation with KYN at week 8 that was not observed atbaseline. Multiple other overlapping correlations in thetwo networks were observed at both baseline and week 8,suggesting that the majority of these interactions were aresult of housekeeping biological interactions and wereprobably not entirely related to the drug effect.

Differential partial correlation networks associated withHRSD17 scores at week 8HRSD17 scores at week 8 indicated the depression status

of the patients post drug treatment. We compared twopartial correlation networks constructed with lower andhigher values of HRSD17 scores at week 8 (the outcomestatus), using a median split, as a node. Our aim was toexamine if the associations between metabolites were

different between patients who responded to the drugbetter than those who responded poorly. Severalmetabolite–metabolite associations across the tyrosine,tryptophan, and purine pathways were found to bechanged as a function of higher or lower outcome status(Supplemental Table 5 A, B). At baseline, GR–MET,TYR–MET, and KYN–URIC partial correlations weremost impacted, while at week 8, KYN–HVA,KYN–3OHKY, 5HTP-G, and HGA–GR values were mostimpacted by HRSD17 week 8 status (Fig. 3e, f). Two sets ofmetabolite–metabolite interactions associated with theoutcome status, HGA–GR interactions at week 8 andMET–TYR interactions at baseline, were further found tobe statistically significant in linear regression models(highlighted in yellow in Fig. 3, e, f). The interaction plotsbased on linear regression models are presented in Sup-plemental Figs. 1 and 2. An interesting observation fromthe differential analysis of networks at baseline was thatthe partial correlations between metabolites that weredifferential between the low versus high HRSD17 networksinvolved several gut-microbe-related metabolites such asHGA, I3AA, 5HIAA, 4HPLA, and 4HPAC amongst oth-ers (Fig. 3e).

Genetic influences on ratios of interacting metabolite pairschange during SSRI treatmentTo identify potential modulators of significant

metabolite–metabolite interactions and their differentialinteractions over time, we performed genome-wideassociation studies with the pairwise ratios of HGA/GRand MET/TYR in 288 subjects at each time point. To thisend, we computed additive genetic associations of the tworatios with 5.55 mio autosomal SNPs at each time point,while adjusting for age, sex, time point-specific HRSD17

score, the first five PCs to account for population strati-fication. The strongest signal for the HGA/GR ratio wasfor rs55933921 on chromosome 7 (baseline: P= 8.59 ×10−7; week 4: P= 3.05 × 10−3; week 8: P= 1.14 × 10−3) ina locus spanning two genes, TAC1 (protachykinin-1) andASNS (asparagine synthetase [glutamine-hydrolyzing]).The strongest signal for the MET/TYR ratio was forrs2701431 on chromosome 15 (baseline: P= 5.57 × 10−3;week 4: P= 2.00 × 10−4; week 8: P= 8.48 × 10−8) in theAGBL1 (ATP/GTP-binding protein like 1) locus (Fig. 4).Of note, genetic associations between these loci andmetabolite ratios were the strongest at the time point thatshowed insignificant metabolite–metabolite interactionson the HRSD17 score.

DiscussionWe have applied a “targeted” electrochemistry-based

metabolomics platform to quantitate the metabolomicprofiles in MDD patients before and after SSRI treatment.Specifically, we assayed 31 neurotransmission-related

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 7 of 14

Fig. 3 Partial correlation networks (PCN). a PCN at baseline; b PCN at baseline after 1000 bootstrap estimations; c PCN at week 8, and d PCN atweek 8 after 1000 bootstrap estimations. The different clusters representing communities of closely associated metabolites are shown in differentcolors. Differential PCN as a function of high versus low HRSD17 week 8 scores at (e) baseline and (f) week 8. The edges between metabolites mostimpacted by higher HRSD17 week 8 scores are bolded in green, while those by lower HRSD17 week 8 scores are bolded in red

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 8 of 14

Fig. 4 Plots showing regional association plots generated with SNiPA61 for: a the HGA/GR ratio at baseline and b the MET/TYR ratio at week 8

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 9 of 14

metabolites that might have some relevance to MDDpathophysiology, based on published literature, includingcompounds from the tryptophan, tyrosine, and purinepathways, in plasma samples from 290 MDD patients atbaseline, 4-, and 8 weeks of escitalopram/citalopramtreatment. Metabolomic profiles were correlated withtreatment response, defined as 50% reduction frombaseline HRSD17 scores. We found that plasma 5HTconcentration was the most significantly decreasedmetabolite among all of the 31 metabolites after the drugtreatment. Higher baseline 5HT levels were associatedwith better response to SSRI treatment. The 5HT levels inresponders remained higher than that those in non-responders all through the treatment period. Comparedwith the baseline metabolic state, significant shifts ofmetabolomic profiles to noncanonical branches of thethree major pathways after drug exposure were noted,such as increase in the production of indoles in TRPmetabolism, phenolic acids in phenylalanine/tyrosinemetabolism, and the PXAN/XAN ratio in the purinepathway. In addition, changes in the catecholaminebranch of tyrosine metabolism like increase in VMA anddecrease in MHPG, and changes in the end products ofpurine pathway, such as decreases in XAN and HX, wereidentified in the MDD patients after drug exposure.Patients who had high MHPG at baseline respondedbetter to SSRI treatment and their MHPG levelsdecreased more significantly than nonresponders. Tem-poral change in serotonin and 5HIAA significantly cor-related with changes in HRSD17 scores over time. Partialcorrelation analysis between metabolites revealed thatMET, TYR, and TRP formed a tight cluster of interactingmolecules in these MDD patients. However, the strengthof interactions (partial correlations) varied significantlypre- and post treatment. GR association with this clusterwas significant at week 8. GWAS for the HGA/GR ratioidentified two genetic loci that mapped to the TAC1 andASNS genes, which are known to be involved in depres-sion and neurotransmission, while the ratio MET/TYRidentified the gene AGBL1 previously linked to neuro-degeneration in mice.Overall, significant perturbations within and between

the tryptophan, tyrosine, and purine pathways due to thedrug exposure, were noted. These findings are consistentwith our previous metabolomic study of sertraline,another SSRI, in depressed patients35, where perturba-tions in TRP, in particular, changes in methoxyindolepathway and the ratio of KYN/TRP were correlated withtreatment outcomes. Interestingly, plasma concentrationsof the indoles synthesized from TRP, I3AA, and I3PA,were found to be significantly increased in the MDDpatients in this study after the SSRI treatment. Plasmaconcentrations for I3AA and I3PA are known to beinfluenced by gut microbiota. Both I3AA and I3PA are

aryl hydrocarbon receptor (AHR) agonists59, which couldactivate AHR transcriptional activity and modulateinflammation in the gut60–63 and brain64,65.Indole-3-propionic acid is a potent hydroxyl radical

scavenger produced exclusively by the commensal gutbacteria Clostridium sporogenes66 and normally found inthe plasma and cerebrospinal fluid. I3AA, on the otherhand, has been found to correlate significantly with bothanxiety and depressive symptoms in chronic kidney dis-ease patients (CKD)67. I3AA can be produced from indoleby gut microflora68 in the intestines, or metabolized intissues from tryptamine69 and other TRP derivatives. Aturemic concentrations, I3AA has been linked to oxidativestress via AHR in CKD patients70. However, neither I3AAnor I3PA changes were correlated to changes in HRSD17

scores in our findings.The most notable change in metabolic profiles after

SSRI exposure occurred in the TRP metabolite, 5HTconcentrations, which was expected from the mechanismof action of SSRI38. Plasma 5HT originates from theenterochromaffin cells in the gut and gets activelyabsorbed and stored by blood platelets, which highlyexpress the 5HT transporter SLC6A4. SSRIs targetSLC6A4 and inhibit 5HT uptake by platelets in blood.Therefore, a dramatic decrease in plasma 5HT con-centration after the SSRI treatment can be expected inthese patients. A higher concentration of plasma 5HT,which was stored in platelets, may reflect an elevatedactivity of the 5HT transporter, and, as a result, greatersensitivity to SSRIs in those patients. This hypotheticalsituation might explain why patients with higher plasma5HT concentrations responded better to SSRIs.Altered metabolic activity of the purine cycle has been

linked with several MDD-related systemic responses, suchas increased proinflammatory and oxidative processes35.The end products of purine metabolism, uric acid, apotent antioxidant, have been reported to be found indecreased levels in MDD71–73, while lower cerebro-spinalfluid (CSF) levels of hypoxanthine and xanthine, the twometabolites preceding uric acid, have previously beenlinked with depression74. Ali-Sisto et al., on the contrary,reported increased levels of xanthine to be associated withMDD75. In our study, we observed higher baseline levelsof xanthine and hypoxanthine that decreased with thedrug treatment. We did not detect increases in uric acid,but we did observe significant increases in the ratios ofparaxanthine/xanthine and xanthosine/xanthine due tothe drug exposure. This may indicate a potential bene-ficial effect of the drug through reducing oxidative stressby direct or indirect inhibition of the xanthine oxidaseenzyme system. XO is known to generate vascular oxi-dative stress through reactive oxygen species productionby catalyzing the hypoxanthine → xanthine → uratesynthesis76. On the other hand, we observed increased

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 10 of 14

associations between uric acid with 4HPLA and also withHGA at week 8 compared with baseline. Uric acid isknown to function as an antioxidant (primarily in plasma)and pro-oxidant (primarily within the cell)77. Para-xanthine showed strong correlations to theophylline at alltime points. It may also be possible that we were not ableto detect significant increases in the levels of the anti-oxidant, uric acid, and the known psychostimulant para-xanthine78 due to their increased turnover rates broughtabout by the drug treatment.In the tyrosine/phenylalanine pathway, 4HPAC and

4HBAC, the phenolic acid metabolites, were found to besignificantly increased in MDD patients after SSRI treat-ment. Although these can potentially come from dietsrich in plant-based foods, evidence suggests that thesecompounds can be produced through microbial fermen-tation of aromatic amino acids (AAAs) in the colon79.Although changes in the concentration of those metabo-lites were not associated with SSRI response in MDDpatients, those changes possibly indicate alterations in gutmicrobiome or gut metabolism after citalopram/escitalo-pram treatment. 4HBAC is known for its antioxidantproperties, as effective scavengers of free radicals andreactive nitrogen species, such as peroxynitrite80. 4HPACis also known for its antioxidant, antiinflammatory, andanticancer activities79.The strong interactions between MET–TYR–TRP,

observed through partial correlation networks at baselineconfirms the connection between folate-mediatedmethionine formation, leading to methyl donation reac-tions that form the monoamine neurotransmitters ser-otonin, dopamine, and epinephrine81,82. In depression,this balance is known to be perturbed83. With the drugexposure, we see further alterations in this balance atweek 8. In addition, we see that GR significantly correlateswith MET and TRP at week 8 post treatment. This may beindicative of changes in methylation status of the ser-otonin transporter84 through epigenetic mechanisms, inresponse to the SSRI treatment in these depressionpatients. At week 8, KYN–3OHKY association decreasedsignificantly with concomitant increases in associationsbetween KYN and the dopamine degradation product,HVA, and this association was comparitively stronger inpatients who did not respond well to the treatment.Using the ratios of metabolites significantly interacting

as intermediate phenotypes leads us to rediscover lociknown to be involved in neurotransmission/depressionand neurodegeneration. The strongest association signalsfor baseline GR/HGA were within a locus on chromo-some 7 containing two central genes: TAC1 (protachy-kinin-1) that has been linked to depression and anxiety85

and ASNS (asparagine synthetase [glutamine-hydrolyz-ing]) that is an important enzyme, the deficiency of whichleads to substantial neurodevelopmental deficits86.

Interestingly, patients with this deficiency (it is an inbornerror of metabolism) also show modest changes in neu-rotransmitters. The strongest signal for MET/TYR waswithin a locus on chromosome 15 containing the geneAGBL1 (ATP/GTP binding protein-like 1) that has a rolein controlling the length of the polyglutamate side chainson tubulin. This process is critical for neuronal survival,and the lack of such control has been reported to result inneurodegeneration in mice87. These findings underscoresthe utility of our “Pharmacometabolomics-Informs-Pharmacogenomics” approach33 to identify candidategenes for further functional studies. Using this strategy,we have previously identified SNP signals in the DEFB1and AHR genes that were associated with severity ofdepressive symptoms in these MDD patients28. DEFB1 isan antimicrobial peptide which is highly expressed andactive in the gut88, playing a potentially important role inmaintaining gut–microbiome homeostasis89. Theseresults fit within the broadening body of information insupport of important roles for the “microbiota–gut–brainaxis” and inflammation in MDD pathophysiology.Several limitations of this study warrant consideration.

Compared with other MDD patients recruited in thePGRN-AMPS trial, study participants were “selected”because they were able to complete all three visits (i.e.,baseline, 4, and 8 weeks) and provide blood samples,which would reduce the number of patients in the finalsample who did poorly. In addition, only Caucasians wereincluded in this study, and thus, a given inherent limita-tion was developed from analyzing a subset of MDDpatients. Furthermore, the LCECA platform capturesinformation on only redox-active compounds in the tyr-osine, tryptophan, purine, and sulfur amino acid pathwaysand several markers of vitamin status and oxidative pro-cesses. The integration of data from lipidomics and massspectrometry-based metabolomics platforms in futurestudies, as well as inclusion of several confounding vari-ables, such as body mass index, diet, and lifestyle factors,would definitely help to better unravel the mechanisticaspect of the drug response.In conclusion, we analyzed the metabolomic profile in

290 MDD patients before and after citalopram/escitalo-pram treatment. Noncanonical metabolic pathways rela-ted to TRP, tyrosine, and purine metabolism were foundto be activated after the drug exposure. There wascrosstalk among these pathways at baseline depressionlevels, which was significantly impacted by the drugexposure. Significant increases in gut–microbiota-relatedmetabolites, such as the indoles and the phenolic acids,were observed in the overall population. Patients whoresponded to the drug compared to those who did not,had significant differences in baseline levels as well as inthe trajectories of several metabolites, including severalgut–microbiota related metabolites, suggesting that the

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 11 of 14

drug exposure might be impacting gut-microbial ecologydifferently in the two groups. Overall, amelioration ofoxidative stress and increases in anti-inflammatory pro-cesses seem to be part of the mechanism involved inresponse to citalopram/escitalopram treatment.

AcknowledgementsThe authors are grateful for the support of NIH, to Lisa Howerton for heradministrative support, and to the study participants and their families of theMayo Pharmacogenomics Research Network-AntidepressantPharmacogenomics Medication Study (PGRN-AMPS). The research and theauthors are supported by funding from the NIH. This work was funded bygrant support to Rima Kaddurah-Daouk through NIH grants R01MH108348,R01AG046171 & U01AG061359, RF1AG051550. Sudeepa Bhattacharyya wassupported by 5R01MH108348, 5R01AG046171-03S1. R.M.W. was supported byNIH grants RO1 GM28157, U19 GM61388, U54 GM114838, and NSF1624615.

Author details1Department of Biomedical Informatics, University of Arkansas for MedicalSciences, Little Rock, AR, USA. 2Department of Psychiatry and Psychology,Mayo Clinic, Rochester, MN, USA. 3Department of Psychiatry and BehavioralSciences, Duke University School of Medicine, Durham, Durham, NC, USA.4Institute of Bioinformatics and Systems Biology, Helmholtz Zentrum München– German Research Center for Environmental Health, Neuherberg, Germany.5Department of Molecular Pharmacology & Experimental Therapeutics, MayoClinic, Rochester, MN, USA. 6Sanofi, Bridgewater, NJ, USA. 7Department ofPsychiatry and Behavioral Sciences, Emory University School of Medicine,Atlanta, GA, USA. 8Department of Psychiatry, Rush University Medical Center,Chicago, IL, USA. 9Texas Tech University, Health Sciences Center, Permian Basin,Odessa, TX, USA. 10Duke-National University of Singapore, Singapore,Singapore. 11Department of Medicine, Duke University, Durham, NC, USA.12Duke Institute of Brain Sciences, Duke University, Durham, NC, USA

Conflict of interestR.M.W. is a cofounder and stockholder in OneOme, LLC, a pharmacogenomicclinical decision support company. A.J.R. has received consulting fees fromAkili, Brain Resource Inc., Compass Inc., Curbstone Consultant LLC., EmmesCorp., Johnson and Johnson (Janssen), Liva-Nova, Mind Linc, Sunovion, and TajMedical; speaking fees from Liva-Nova; and royalties from Guilford Press andthe University of Texas Southwestern Medical Center, Dallas, TX (for theInventory of Depressive Symptoms and its derivatives). He is also named thecoinventor on two patents: US Patent No. 7,795,033: Methods to Predict theOutcome of Treatment with Antidepressant Medication and US Patent No.7,906,283: Methods to Identify Patients at Risk of Developing Adverse Eventsduring Treatment with Antidepressant Medication. M.A.F. has received grantsupport from AssureRx Health Inc, Myriad, Pfizer Inc, NIMH (R01 MH079261),the National Institute on Alcohol Abuse and Alcoholism (P20AA017830) in theNational Institutes of Health at the US Department of Health and HumanServices, and the Mayo Foundation. He has been a consultant (for Mayo) toJanssen Global Services, LLC; Mitsubishi Tanabe Pharma Corp; Myriad Genetics,Inc; Sunovion Pharmaceuticals, Inc; and Teva Pharmaceutical Industries Ltd. Hehas received continuing medical education, travel, and presentation supportfrom the American Physician Institute and CME Outfitters. L.W. was supportedby NIH grants U19 GM61388, U54 GM114838, and NSF164615. She is acofounder and stockholder in OneOme. R.K.-D. is an inventor on key patents inthe field of metabolomics. D.N.’s stipend has been supported in part by NIHT32 GM072474 and the Mayo Graduate School. A.T.A.’s research was supportedby National Institute of General Medical Sciences of the National Institutes ofHealth under award number T32 GM008685. M.A. was supported by theNational Institute on Aging [R01AG057452, RF1AG051550, and R01AG046171],National Institute of Mental Health [R01MH108348], and Qatar NationalResearch Fund [NPRP8-061-3-011]. The funders listed above had no role in thedesign and conduct of the study; collection, management, analysis, andinterpretation of the data; preparation, review, or approval of the paper; anddecision to submit the manuscript for publication.

Publisher’s noteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Supplementary Information accompanies this paper at (https://doi.org/10.1038/s41398-019-0507-5).

Received: 5 September 2018 Revised: 29 March 2019 Accepted: 29 April2019

References1. World Health Organization W. Depression and Other Common Mental Disorders:

Global Health Estimates. (World Health Organization, Geneva, 2017).2. Crismon, M. L. et al. The Texas Medication Algorithm Project: report of the

Texas Consensus Conference Panel on medication treatment of majordepressive disorder. J. Clin. Psychiatry 60, 142–156 (1999).

3. Anderson, I. M. et al. Evidence-based guidelines for treating depressive dis-orders with antidepressants: a revision of the 2000 British Association forPsychopharmacology guidelines. J. Psychopharmacol. 22, 343–396 (2008).

4. Rush, A. J. et al. Report by the ACNP Task Force on response and remission inmajor depressive disorder. Neuropsychopharmacol.: Off. Publ. Am. Coll. Neu-ropsychopharmacol. 31, 1841–1853 (2006).

5. Rush, A. J. et al. Selecting among second-step antidepressant medicationmonotherapies: predictive value of clinical, demographic, or first-step treat-ment features. Arch. Gen. Psychiatry 65, 870–880 (2008).

6. US Food and Drug Administration Ws. Antidepressant Use in Children,Adolescents, and Adults. (2007).

7. Rush, A. J. et al. Acute and longer-term outcomes in depressed outpatientsrequiring one or several treatment steps: a STAR*D report. Am. J. Psychiatry163, 1905–1917 (2006).

8. Kaddurah-Daouk, R., Kristal, B. S. & Weinshilboum, R. M. Metabolomics: a globalbiochemical approach to drug response and disease. Annu. Rev. Pharmacol.Toxicol. 48, 653–683 (2008).

9. Kaddurah-Daouk, R. & Krishnan, K. R. Metabolomics: a global biochemicalapproach to the study of central nervous system diseases. Neuropsycho-pharmacol.: Off. Publ. Am. Coll. Neuropsychopharmacol. 34, 173–186 (2009).

10. Kaddurah-Daouk, R. & Weinshilboum, R. M. Pharmacometabolomics: implica-tions for clinical pharmacology and systems pharmacology. Clin. Pharmacol.Ther. 95, 154–167 (2014).

11. Neavin, D., Kaddurah-Daouk, R. & Weinshilboum, R. Pharmacometabolomicsinforms pharmacogenomics. Metab.: Off. J. Metab. Soc. 12, 121 (2016).

12. Gadad, B. S. et al. Peripheral biomarkers of major depression and anti-depressant treatment response: current knowledge and future outlooks. J.Affect. Disord. 233, 3–14 (2018).

13. Rush, A. J. & Ibrahim, H. M. A clinician’s perspective on biomarkers. Focus 16,124–134 (2018).

14. Paige, L. A., Mitchell, M. W., Krishnan, K. R., Kaddurah-Daouk, R. & Steffens, D. C.A preliminary metabolomic analysis of older adults with and withoutdepression. Int. J. Geriatr. Psychiatry 22, 418–423 (2007).

15. Miller, C. L., Llenos, I. C., Dulay, J. R. & Weis, S. Upregulation of the initiating stepof the kynurenine pathway in postmortem anterior cingulate cortex fromindividuals with schizophrenia and bipolar disorder. Brain Res. 1073–1074,25–37 (2006).

16. Myint, A. M. et al. Tryptophan breakdown pathway in bipolar mania. J. Affect.Disord. 102, 65–72 (2007).

17. Holmes, E. et al. Metabolic profiling of CSF: evidence that early interventionmay impact on disease progression and outcome in schizophrenia. PLoS Med.3, e327 (2006).

18. Kaddurah-Daouk, R. et al. Metabolomic mapping of atypical antipsychoticeffects in schizophrenia. Mol. Psychiatry 12, 934–945 (2007).

19. Tkachev, D., Mimmack, M. L., Huffaker, S. J., Ryan, M. & Bahn, S. Further evi-dence for altered myelin biosynthesis and glutamatergic dysfunction inschizophrenia. Int. J. Neuropsychopharmacol. 10, 557–563 (2007).

20. Hayaishi, O. Utilization of superoxide anion by indoleamine oxygenase-catalyzed tryptophan and indoleamine oxidation. Adv. Exp. Med. Biol. 398,285–289 (1996).

21. Maes, M. et al. Treatment with interferon-alpha (IFN alpha) of hepatitis Cpatients induces lower serum dipeptidyl peptidase IV activity, which is related

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 12 of 14

to IFN alpha-induced depressive and anxiety symptoms and immune acti-vation. Mol. Psychiatry 6, 475–480 (2001).

22. Stone, T. W. & Darlington, L. G. Endogenous kynurenines as targets for drugdiscovery and development. Nat. Rev. Drug Discov. 1, 609–620 (2002).

23. Wirleitner, B., Neurauter, G., Schrocksnadel, K., Frick, B. & Fuchs, D. Interferon-gamma-induced conversion of tryptophan: immunologic and neuropsychia-tric aspects. Curr. Med. Chem. 10, 1581–1591 (2003).

24. Schiepers, O. J. G., Wichers, M. C. & Maes, M. Cytokines and major depression.Prog. Neuro-Psychopharmacol. Biol. Psychiatry 29, 201–217 (2005).

25. Gabbay, V. et al. The possible role of the kynurenine pathway in adolescentdepression with melancholic features. J. Child Psychol. Psychiatry, Allied Discip.51, 935–943 (2010).

26. Bradley, K. A. L. et al. The role of the kynurenine pathway in suicidality inadolescent major depressive disorder. Psychiatry Res. 227, 206–212 (2015).

27. Reus, G. Z. et al. Kynurenine pathway dysfunction in the pathophysiology andtreatment of depression: evidences from animal and human studies. J. Psy-chiatr. Res. 68, 316–328 (2015).

28. Liu, D. et al. Beta-defensin 1, aryl hydrocarbon receptor and plasma kynureninein major depressive disorder: metabolomics-informed genomics. Transl. Psy-chiatry 8, 10 (2018).

29. Park, D. I. et al. Purine and pyrimidine metabolism: convergent evidence onchronic antidepressant treatment response in mice and humans. Sci. Rep. 6,35317 (2016).

30. Kaddurah-Daouk, R. et al. Cerebrospinal fluid metabolome in mood disorders-remission state has a unique metabolic profile. Sci. Rep. 2, 667 (2012).

31. Kaddurah-Daouk, R. & Weinshilboum, R. Metabolomic signatures for drugresponse phenotypes: pharmacometabolomics enables precision medicine.Clin. Pharmacol. Ther. 98, 71–75 (2015).

32. Wikoff, W. R. et al. Pharmacometabolomics reveals racial differences inresponse to atenolol treatment. PLoS ONE 8, e57639 (2013).

33. Yerges-Armstrong, L. M. et al. Purine pathway implicated in mechanism ofresistance to aspirin therapy: pharmacometabolomics-informed pharmaco-genomics. Clin. Pharmacol. Ther. 94, 525–532 (2013).

34. Zhu, H. et al. Pharmacometabolomics of response to sertraline and to placeboin major depressive disorder—possible role for methoxyindole pathway. PLoSONE 8, e68283 (2013).

35. Kaddurah-Daouk, R. et al. Pretreatment metabotype as a predictor of responseto sertraline or placebo in depressed outpatients: a proof of concept. Transl.Psychiatry 1, e26 (2011).

36. Rotroff, D. M. et al. Metabolomic signatures of drug response phenotypes forketamine and esketamine in subjects with refractory major depressive dis-order: new mechanistic insights for rapid acting antidepressants. Transl. Psy-chiatry 6, e894 (2016).

37. Ji, Y. et al. Glycine and a glycine dehydrogenase (GLDC) SNP as citalo-pram/escitalopram response biomarkers in depression:pharmacometabolomics-informed pharmacogenomics. Clin. Pharmacol.Ther. 89, 97–104 (2011).

38. Gupta, M. et al. TSPAN5, ERICH3 and selective serotonin reuptake inhibitors inmajor depressive disorder: pharmacometabolomics-informed pharmacoge-nomics. Mol. Psychiatry 21, 1717–1725 (2016).

39. Mrazek, D. A. et al. Treatment outcomes of depression: the pharmacogenomicresearch network antidepressant medication pharmacogenomic study. J. Clin.Psychopharmacol. 34, 313–317 (2014).

40. Ji, Y. et al. Citalopram and escitalopram plasma drug and metabolite con-centrations: genome-wide associations. Br. J. Clin. Pharmacol. 78, 373–383(2014).

41. Ji, Y. et al. Pharmacogenomics of selective serotonin reuptake inhibitortreatment for major depressive disorder: genome-wide associations andfunctional genomics. Pharm. J. 13, 456–463 (2013).

42. Rush, A. J. et al. Bupropion-SR, sertraline, or venlafaxine-XR after failure of SSRIsfor depression. N. Engl. J. Med. 354, 1231–1242 (2006).

43. Matson, W. R. et al. n-Electrode three-dimensional liquid chromatography withelectrochemical detection for determination of neurotransmitters. Clin. Chem.30, 1477–1488 (1984).

44. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practicaland powerful approach to multiple testing. J. R. Stat. Soc. Ser. B (Methodol.) 57,289–300 (1995).

45. Pons, P. & Latapy, M. Computing communities in large networks using ran-dom walks. In Proc. 20th International Conference on Computer and InformationSciences. 284–293 (Springer-Verlag, Istanbul, Turkey, 2005).

46. von Elm, E. et al. The Strengthening the Reporting of Observational Studies inEpidemiology (STROBE) statement: guidelines for reporting observationalstudies. Ann. Intern. Med. 147, 573–577 (2007).

47. Howie, B. N., Donnelly, P. & Marchini, J. A flexible and accurate genotypeimputation method for the next generation of genome-wide associationstudies. PLoS Genet. 5, e1000529 (2009).

48. Delaneau, O., Zagury, J. F. & Marchini, J. Improved whole-chromosomephasing for disease and population genetic studies. Nat. Methods 10, 5–6(2013).

49. Chang, C. C. et al. Second-generation PLINK: rising to the challenge of largerand richer datasets. GigaScience 4, 7 (2015).

50. Golino, H. F. & Epskamp, S. Exploratory graph analysis: a new approach forestimating the number of dimensions in psychological research. PLoS ONE 12,e0174035 (2017).

51. Do, K. T. et al. Characterization of missing values in untargeted MS-basedmetabolomics data and evaluation of missing data handling strategies.Metabol.: Off. J. Metabol. Soc. 14, 128 (2018).

52. Pinheiro J. C., Bates D. J. D., DebRoy S., Sakar D. The Nlme Package: Linear andNonlinear Mixed Effects Models, R Version 3, vol. 6 (2012).

53. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate – a practicaland powerful approach to multiple testing. J. R. Stat. Soc. B Met. 57, 289–300(1995).

54. Conesa, A1., Nueda, M. J., Ferrer, A. & Talón, M. maSigPro: a method to identifysignificantly differential expression profiles in time-course microarray experi-ments. Bioinformatics 22, 1096–1102 (2006).

55. van Borkulo C. D., Jonas H., Epskamp S., Millner A. Network Comparison Test:statistical comparison of two networks based on three invariance measures (Rpackage Version 2.0.1) [Computer Software]. (2016).

56. Auton, A. et al. A global reference for human genetic variation. Nature 526,68–74 (2015).

57. Krumsiek, J., Suhre, K., Illig, T., Adamski, J. & Theis, F. J. Gaussian graphicalmodeling reconstructs pathway reactions from high-throughput metabo-lomics data. BMC Syst. Biol. 5, 21 (2011).

58. Shin, S.-Y. et al. An atlas of genetic influences on human blood metabolites.Nat. Genet. 46, 543 (2014).

59. Hubbard, T. D., Murray, I. A. & Perdew, G. H. Indole and tryptophan metabo-lism: endogenous and dietary routes to Ah receptor activation. Drug Metab.Dispos. 43, 1522–1535 (2015).

60. Whitfield-Cargile, C. M. et al. The microbiota-derived metabolite indoledecreases mucosal inflammation and injury in a murine model of NSAIDenteropathy. Gut Microbes 7, 246–261 (2016).

61. Tilg, H. Diet and intestinal immunity. N. Engl. J. Med. 366, 181–183 (2012).62. Li, Y. et al. Exogenous stimuli maintain intraepithelial lymphocytes via aryl

hydrocarbon receptor activation. Cell 147, 629–640 (2011).63. Schiering, C. et al. Feedback control of AHR signalling regulates intestinal

immunity. Nature 542, 242–245 (2017).64. Rothhammer, V. et al. Type I interferons and microbial metabolites of tryp-

tophan modulate astrocyte activity and central nervous system inflammationvia the aryl hydrocarbon receptor. Nat. Med. 22, 586–597 (2016).

65. Rothhammer, V. et al. Microglial control of astrocytes in response to microbialmetabolites. Nature 557, 724–728 (2018).

66. Rosas, H. D. et al. A systems-level “misunderstanding”: the plasma metabolomein Huntington’s disease. Ann. Clin. Transl. Neurol. 2, 756–768 (2015).

67. Karu, N. et al. Tryptophan metabolism, its relation to inflammation and stressmarkers and association with psychological and cognitive functioning: Tas-manian Chronic Kidney Disease pilot study. BMC Nephrol. 17, 171 (2016).

68. Evenepoel, P., Meijers, B. K., Bammens, B. R. & Verbeke, K. Uremic toxins ori-ginating from colonic microbial metabolism. Kidney Int. Suppl. 114, S12–S19(2009).

69. Coppen, A., Brooksbank, B. W., Eccleston, E., Peet, M. & White, S. G. Tryptophanmetabolism in depressive illness. Psychol. Med. 4, 164–173 (1974).

70. Sallee, M. et al. The aryl hydrocarbon receptor-activating effect of uremictoxins from tryptophan metabolism: a new concept to understand cardio-vascular complications of chronic kidney disease. Toxins 6, 934–949 (2014).

71. Chaudhari, K. et al. Clinical correlation of alteration of endogenous antioxidant-uric acid level in major depressive disorder. Indian J. Clin. Biochem.: IJCB 25,77–81 (2010).

72. Kesebir, S., Tatlidil Yaylaci, E., Suner, O. & Gultekin, B. K. Uric acid levels may be abiological marker for the differentiation of unipolar and bipolar disorder: therole of affective temperament. J. Affect. Disord. 165, 131–134 (2014).

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 13 of 14

73. Wen, S. et al. Serum uric acid levels and the clinical characteristics ofdepression. Clin. Biochem. 45, 49–53 (2012).

74. Agren, H., Niklasson, F. & Hallgren, R. Brain purinergic activity linked withdepressive symptomatology: hypoxanthine and xanthine in CSF of patientswith major depressive disorders. Psychiatry Res. 9, 179–189 (1983).

75. Ali-Sisto, T. et al. Purine metabolism is dysregulated in patients with majordepressive disorder. Psychoneuroendocrinology 70, 25–32 (2016).

76. George, J. & Struthers, A. D. Role of urate, xanthine oxidase and the effects ofallopurinol in vascular oxidative stress. Vasc. Health Risk Manag. 5, 265–272(2009).

77. Sautin, Y. Y. & Johnson, R. J. Uric acid: the oxidant–antioxidant paradox.Nucleosides, Nucleotides Nucleic Acids 27, 608–619 (2008).

78. Orru, M. et al. Psychostimulant pharmacological profile of paraxanthine, themain metabolite of caffeine in humans. Neuropharmacology 67, 476–484(2013).

79. Russell, W. R. et al. Major phenylpropanoid-derived metabolites in the humangut can arise from microbial fermentation of protein. Mol. Nutr. Food Res. 57,523–535 (2013).

80. Hubkova, B., Velika, B., Birkova, A., Guzy, J. & Marekova, M. Hydroxybenzoic acidsand their derivatives as peroxynitrite scavengers. Free Radic. Biol. Med. 75(Suppl1), S33–S34 (2014).

81. Weinshilboum, R. M., Raymond, F. A., Elveback, L. R. & Weidman, W. H. Cor-relation of erythrocyte catechol-O-methyltransferase activity between siblings.Nature 252, 490–491 (1974).

82. Weinshilboum, R. M., Otterness, D. M. & Szumlanski, C. L. Methylation phar-macogenetics: catechol O-methyltransferase, thiopurine methyltransferase,and histamine N-methyltransferase. Annu. Rev. Pharmacol. Toxicol. 39, 19–52(1999).

83. Miller, A. L. The methylation, neurotransmitter, and antioxidant connectionsbetween folate and depression. Altern. Med. Rev. 13, 216–226 (2008).

84. Shi, M. et al. Methylation status of the serotonin transporter promoter CpGisland is associated with major depressive disorder in Chinese Han population:a case-control study. J. Nerv. Ment. Dis. 205, 641–646 (2017).

85. Bilkei-Gorzo, A., Racz, I., Michel, K. & Zimmer, A. Diminished anxiety- anddepression-related behaviors in mice with selective deletion of the Tac1 gene.J. Neurosci.: Off. J. Soc. Neurosci. 22, 10046–10052 (2002).

86. Ruzzo, E. K. et al. Deficiency of asparagine synthetase causes congenitalmicrocephaly and a progressive form of encephalopathy. Neuron 80, 429–441(2013).

87. Cocchi, E. et al. Genome-wide association study of antidepressant response:involvement of the inorganic cation transmembrane transporter activitypathway. BMC Psychiatry 16, 106 (2016).

88. Schroeder, B. O. et al. Reduction of disulphide bonds unmasks potentantimicrobial activity of human beta-defensin 1. Nature 469, 419–423(2011).

89. Ostaff, M. J., Stange, E. F. & Wehkamp, J. Antimicrobial peptides and gutmicrobiota in homeostasis and pathology. EMBO Mol. Med. 5, 1465–1483(2013).

Bhattacharyya et al. Translational Psychiatry (2019) 9:173 Page 14 of 14